ofz#20455 sw: HTML import: fix invalid table in footer

The obvious problem was that a bookmark failed to be copied to the
correct node, it was created on a SwStartNode, which failed in
makeMark() and caused a null-pointer.

The target position was off by 1 node because there was a spurious
StartNode/EndNode pair directly below the table:

[  41]   0x5b13430          TableNode ,
[  42]    0x5b1d010         StartNode ,
[  43]     0x5b12a50        StartNode ,
[  44]      0x5b135f0        TextNode "",

This was created by a special case in SwTableNode::MakeCopy() because
.GetTabSortBoxes().size() == 1.

But the table had actually quite a bunch more cells in the nodes-array,
just they were not yet in the SwTable.

In an exciting twist of events, it turns out the table was copied while
it was not yet finished parsing: the problem was that in the middle of
the table, some CSS set some page attributes, and this caused a
first-page page style to be created in SwCSS1Parser::ParseStyleSheet(),
by copying the master page style.

Unfortunately the table was in the <div title="footer">, so it was
copied in this incomplete and inconsistent state.

It might be possible to get rid of the special case in
SwTableNode::MakeCopy() by restricting the special case skipping of
StartNodes at the start in SwNodes::CopyNodes() a bit so that StartNodes
whose EndNodes are copied aren't skipped; at least that's the most
reasonable explanation for the special case.

But for now just fix the HTML import.

Additionally, only on MacOSX, using libc++, this triggered an assert:
  Assertion failed: (!pImpl->mpStaticDefaults || typeid(rItem) == typeid(GetDefaultItem(nWhich))), function PutImpl, file /Users/tdf/lode/jenkins/workspace/lo_gerrit/Config/macosx_clang_dbgutil/svl/source/items/itempool.cxx, line 611.

Probably because SdrTextAniCountItem is not marked DLLPUBLIC.

Change-Id: Ia167265e7540eea649801eaac2b89f9e18b685cd
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/87859
Tested-by: Jenkins
Reviewed-by: Michael Stahl <michael.stahl@cib.de>

1 files changed, 13 insertions, 5 deletions

diff --git a/sw/source/filter/html/htmlcss1.cxx b/sw/source/filter/html/htmlcss1.cxx
index ca28eac168b5..c90c4e73307a 100644
--- a/sw/source/filter/html/htmlcss1.cxx
+++ b/sw/source/filter/html/htmlcss1.cxx
@@ -87,11 +87,13 @@ void SwCSS1Parser::ChgPageDesc( const SwPageDesc *pPageDesc,
         m_pDoc->ChgPageDesc( pos, rNewPageDesc );
 }
 
-SwCSS1Parser::SwCSS1Parser( SwDoc *pD, const sal_uInt32 aFHeights[7], const OUString& rBaseURL, bool bNewDoc ) :
-    SvxCSS1Parser( pD->GetAttrPool(), rBaseURL,
-                   aItemIds, SAL_N_ELEMENTS(aItemIds)),
-    m_pDoc( pD ),
-    m_nDropCapCnt( 0 ),
+SwCSS1Parser::SwCSS1Parser(SwDoc *const pDoc, SwHTMLParser const& rParser,
+        const sal_uInt32 aFHeights[7], const OUString& rBaseURL, bool const bNewDoc)
+    : SvxCSS1Parser(pDoc->GetAttrPool(), rBaseURL,
+                   aItemIds, SAL_N_ELEMENTS(aItemIds))
+    , m_pDoc( pDoc )
+    , m_rHTMLParser(rParser)
+    , m_nDropCapCnt( 0 ),
     m_bIsNewDoc( bNewDoc ),
     m_bBodyBGColorSet( false ),
     m_bBodyBackgroundSet( false ),
@@ -1341,6 +1343,12 @@ const SwPageDesc *SwCSS1Parser::GetPageDesc( sal_uInt16 nPoolId, bool bCreate )
     const SwPageDesc *pPageDesc = FindPageDesc(m_pDoc, nPoolId);
     if( !pPageDesc && bCreate )
     {
+        if (m_rHTMLParser.IsReadingHeaderOrFooter())
+        {   // (there should be only one definition of header/footer in HTML)
+            SAL_WARN("sw.html", "no creating PageDesc while reading header/footer");
+            return nullptr;
+        }
+
         // The first page is created from the right page, if there is one.
         SwPageDesc *pMasterPageDesc = nullptr;
         if( RES_POOLPAGE_FIRST == nPoolId )